Guide for library design and bias correction for large-scale transcriptome studies using highly multiplexed RNAseq...

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Standard RNA sequencing (RNAseq) methods, including bulk and single-cell RNAseq, utilize DNA barcodes for sample and cell identification, pooling barcoded cDNAs before sequencing.
• High-throughput sequencing often involves multiple library pools for large studies, but inter-library correlation can be low due to batch effects, hindering data integration.
• Single-cell and low-input RNAseq methods amplify pooled libraries via PCR, potentially introducing further biases.

Purpose of the Study:

• To address the challenge of low correlation and batch effects between multiple RNAseq library pools.
• To develop and validate a computational method for correcting library biases in highly multiplexed sequencing data.
• To improve the reliability and integration of gene expression profiles from large-scale RNAseq studies.

Main Methods:

• Investigated 166 technical replicates across 14 RNAseq libraries prepared using the STRT method.
• Developed and applied the NBGLM-LBC (Negative Binomial Generalized Linear Model - Library Bias Correction) algorithm.
• Utilized simulation experiments to assess the requirements for effective library bias correction, such as consistent sample layout.

Main Results:

• Identified gene-specific patterns in library biases and their association with uneven library yields.
• Successfully corrected gene expression biases using the NBGLM-LBC algorithm.
• Demonstrated that while low library yields couldn't be directly corrected, omitting low-yield libraries resolved the issue; NBGLM-LBC requires a consistent sample layout for optimal performance.

Conclusions:

• The NBGLM-LBC algorithm effectively resolves library biases in RNAseq data.
• The R source code for NBGLM-LBC is publicly available, facilitating its application in various research fields.
• This method is suitable for studies employing highly multiplexed sequencing with consistent sample distribution across libraries, such as case-control comparisons.